def test_getValues(self):
data = pd.DataFrame({
'r1': np.arange(5),
'r2': np.arange(5)**2,
'r3': np.arange(5)**3
}).T
binMask = [[True, True, False, False, True],
[True, True, False, False, False],
[False, True, True, True, True]]
maskedData = MaskedArray(data=data, mask=binMask)
res_rows = [masked for masked in maskedData.getMasked(rows=True)]
self.assertEqual(res_rows, [[2, 3], [4, 9, 16], [0]])
res_cols = [masked for masked in maskedData.getMasked(rows=False)]
self.assertEqual(res_cols, [[0], [], [2, 4], [3, 9], [16]])
def score_model(model, data, metric, cols=None):
# Create masked array
if cols is None:
cols = data.columns
maskedData = MaskedArray(data=data)
maskedData.generate()
maskedDf = pd.DataFrame(maskedData.getMaskedMatrix(),
index=data.index,
columns=data.columns)
# Predict
# model.fit(maskedDf)
imputed = model.predict(maskedDf)
imputedGenes = np.intersect1d(cols, imputed.columns)
# Compare imputed masked array and input
maskedIdx = maskedDf[imputedGenes].values != data[imputedGenes].values
score_res = metric(data[imputedGenes].values[maskedIdx],
imputed[imputedGenes].values[maskedIdx])
return score_res
def test_generate(self):
rawData = test_data.rawData
m_df = MaskedArray(data=rawData, dropout=0.1)
m_df.generate()